Quoting Chunming Zhou (2018-10-21 12:14:24) > v2: > add a mutex between sync_cb execution and free. The result would appear to be that syncobj->lock is relegated to protecting the pt_list and the mutex would only be needed for the syncobj->cb_list. The patch looks correct for resolving the deadlock and avoiding introducing any new fails, but I do wonder if spinlock_t pt_lock; struct mutex cb_mutex; and clearly separating the roles would not be a better approach. -Chris _______________________________________________ dri-devel mailing list dri-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/dri-devel